Spotting AI-Generated Salary Slips: 7 Signs

An AI-generated salary slip is a synthetic PDF produced by a language model or a template-driven generator that mimics the layout of a payroll system's output. The visual fidelity is now high enough that font, alignment, and totals are usually correct. Where AI fabrications break is the cross-system footprint a real salary slip leaves behind: a Form 26AS entry for the TDS deposit, an Annual Information Statement (AIS) line for the salary credit, a Universal Account Number (UAN) recording the EPF contribution, and a Ministry of Corporate Affairs registry record for the employer entity.

Seven surface-level signs surface on the document itself, three institutional cross-checks confirm or refute what the document claims, and one verification QR short-circuits the whole verification loop when present. For HR teams processing Form 12BB submissions, for credit officers running underwriting on home and personal loans, and for CA practices reviewing client filings, the workflow below catches the bulk of AI fabrications in under three minutes per file.

The stakes for the verifier are real. Under Section 192 (now Section 392 under the Income-tax Act, 2025 for Tax Year 2026-27 onwards), the employer is liable for any TDS shortfall that arises from accepting an unsupported HRA exemption claim. Under Section 201(1A), interest accrues from the date the TDS should have been deducted; Section 271C adds a penalty equal to the shortfall. Lenders carry their own write-off risk plus regulatory scrutiny under the RBI Master Direction on KYC. The cost of a missed AI fabrication is borne by the institution, not the applicant.

A real payroll system computes net pay against a complicated arithmetic stack: basic, allowances, deductions for Section 17 perquisites, TDS at the slab rate applied month-by-month, EPF at 12 percent of basic (with a Rs. 1,800 statutory floor and a Rs. 15,000 wage ceiling for the standard category), ESI at 0.75 percent of gross for eligible employees, and professional tax that varies by state. The arithmetic produces a number with paise.

Real employer payroll software either rounds to the nearest rupee (most cases) or carries paise on the slip and rounds at credit time. The variation across months is meaningful: holiday adjustments, leave-without-pay days, mid-month joiners, mid-year salary revisions, and TDS rounding all push the net-pay figure by Rs. 50 to Rs. 500 month over month at a Rs. 60,000 net-pay level.

AI-generated slips frequently produce net pay that is identical to the rupee across three or four months. The pattern surfaces because the language model is given a single net-pay target by the user and reproduces it month over month. The cross-check is straightforward: pull the applicant's bank statement for the same three months and look for matching credits. Genuine credits vary by Rs. 100 to Rs. 500. Identical credits to the rupee across three consecutive months at any salary band above Rs. 50,000 is a hard flag.

The Universal Account Number is a 12-digit identifier issued by the Employees Provident Fund Organisation under the EPF Act. Any employer with 20 or more employees is statutorily required to register with EPFO and to issue a UAN to every covered employee. The UAN appears on the salary slip alongside the PF deduction line.

Three red flags appear here. First, the UAN field is missing entirely on the slip. Second, the UAN is present but the format is wrong: real UANs are exactly 12 digits, beginning with a numeric leader code; AI fabrications occasionally produce alphanumeric strings or shorter numbers. Third, the UAN is well-formatted but does not resolve on the EPFO member portal — a verifier can paste the UAN into the public lookup and confirm whether a member record exists.

For employers under the 20-employee threshold, EPFO registration is optional, so the UAN may be genuinely absent. The verifier checks whether the employer is below the threshold by looking up the company on the MCA registry. A 200-employee company without a UAN on its slip is suspicious; a 12-employee proprietorship without a UAN is fine. For applicants whose income proof reconciles partially against EPFO, the salary-slip and Form 16 reconciliation procedure isolates the missing element.

Every company incorporated in India carries a Corporate Identification Number (CIN) on the MCA registry. Private limited companies, public companies, LLPs, and one-person companies all appear there. Sole proprietorships and partnerships outside the LLP form do not appear on MCA, but they will appear on the GST registry (gst.gov.in) if their turnover crosses the registration threshold or if they voluntarily registered.

A verifier runs the employer name through MCA master data and checks two outputs: whether the company exists at all, and whether its status is Active. A struck-off, dormant, or under-process-of-strike-off company has a CIN that resolves but a status flag that breaks the assumption of ongoing payroll. AI fabrications often produce plausible-sounding employer names (Acme Technologies Private Limited, Pinnacle Solutions India Limited) that do not exist on MCA.

The sister check is the employer's GSTIN (if listed on the slip — many slips do carry it). A GSTIN is 15 characters in the AAAAAGGGGNNZNN format and can be validated on the GSTN public portal using the search-taxpayer feature. A malformed GSTIN, an inactive GSTIN, or a GSTIN that resolves to a different entity than the slip's listed name are all flags. For self-employed applicants and freelancers whose income proof relies on GST registration data, the non-GST invoice format and bill of supply framework explains what valid alternatives look like.

Real salary slips carry an authorised-signatory block: a name, a designation, and a signature or digital stamp. The designation lines up with the entity type. A private limited company's slips carry a signatory whose designation is Director, CFO, HR Head, or Authorised Representative; a partnership's slips carry a Partner; a sole proprietorship carries the Proprietor or an authorised Manager.

AI fabrications routinely produce internally inconsistent signatory blocks. A slip purportedly issued by a Rs. 200-crore-revenue private limited company carries a signatory designated Proprietor (legally impossible — a private limited company has Directors, not Proprietors). A partnership firm's slip carries a Director. A small proprietorship carries a CFO. The mismatch surfaces in five seconds when the verifier reads the signatory line against the entity name.

The sister check: where the slip carries a signatory name, is the named person on the MCA Director and Key Managerial Personnel list for that company? Most private companies have 2-5 directors and 1-3 KMP. The names appear publicly on MCA. If the slip names Mr. A as Director but MCA shows the Directors as Mr. B and Mr. C, the slip is fabricated or out of date by months. For the parallel rent-side check on landlord identity, the Rule 26C employer HRA verification checklist covers the analogous lookup on the landlord PAN.

This is the single highest-signal verification step. Form 26AS is a PAN-linked statement maintained by the Income Tax Department that lists every TDS deposit made against the applicant's PAN by every deductor, quarter by quarter. The applicant can pull their own Form 26AS from the e-Filing portal and submit it; HR teams and lenders re-pull through the same portal where they have legitimate consent.

When the salary slip claims a TDS deduction of, say, Rs. 12,000 per month, three months of slips should correspond to Rs. 36,000 of TDS visible in Form 26AS under the same employer's TAN by the end of the relevant quarter. The TAN on the slip should match the TAN on the 26AS entry. The cumulative TDS, the quarter, and the deductor name should align.

Three failure modes flag the slip. First, no entry at all in 26AS — the slip is fabricated or the employer has defaulted on TDS deposit. Second, an entry exists but at a different TAN — the slip's employer is misrepresented. Third, the cumulative TDS on 26AS is materially lower than the slip's running total — the slip is overstating either the deduction rate or the salary. Any of the three is grounds to reject the slip as supporting evidence for a credit decision or an HRA submission.

The sister statement is the AIS. While 26AS focuses on TDS, the AIS aggregates salary credits reported by the employer alongside other Specified Financial Transactions. A salary slip claiming Rs. 18 lakh annual gross should align with the AIS salary line within a 2-3 percent variance. Material divergence is a hard rejection.

Every salary structure produces a recognisable bank-credit pattern. Salaried employees in regular employment receive a credit on a fixed day of the month (1st, 5th, last working day, mid-month) from a transferor name that matches the employer (often abbreviated in the bank narration). Contract employees may receive credits on irregular days. Variable-pay employees see split credits for fixed salary and quarterly bonus.

AI-generated slips are often submitted alongside bank statements that the applicant has provided. The cross-check is mechanical: every monthly slip in the submission set should correspond to a credit in the bank statement within three to five days of the slip's pay-period end, in an amount matching the slip's net pay (within rounding). When the slips show 12 months at Rs. 1,20,000 net but the bank statement shows credits varying between Rs. 80,000 and Rs. 1,40,000 month over month from a transferor name that doesn't match the employer, the slip pack and the bank statement are not telling the same story.

The extreme case: slips submitted without a corresponding bank statement. Where the applicant claims they receive salary in cash, the verifier should treat the slip as unverifiable income evidence and ask for the Form 16 (which the employer must issue under Section 192 for any salary above the threshold) and the corresponding 26AS entry. Cash salary is not in itself a flag — many SMB employees receive cash and rely on Form 16 for the formal evidence — but the absence of any bank-side trace shifts the entire verification burden to the tax-side cross-checks.

The visual tells are the least reliable signal in 2026 because AI image generation has caught up to template fidelity. They still surface on lower-effort fabrications, and they are the fastest to scan. Five recurring patterns:

• Font inconsistency within a single field. The Rs. symbol uses a different font from the amount; one column's header uses Arial while the body uses Helvetica. Real payroll-system output uses one font family per slip.

• Vertical misalignment in the deductions table. The PF figure sits half a line above the ESI figure; the column gridlines do not pass cleanly through both. Real tables have rigid cell alignment.

• Watermark layered above the text instead of behind it. A real watermark appears at low opacity behind the printed content. AI-rendered watermarks sometimes appear at high opacity above the content, partially obscuring numbers.

• Logo resolution mismatch with the company's actual logo. The fabricator pulled the logo from a Google image search at a different resolution than the company's brand-asset version. Pixelation, white-edge artefacts, and aspect-ratio distortion give it away.

• Date format inconsistency. The pay-period field reads 01-Apr-2026 while the issue-date field reads 4/30/2026. Genuine payroll output uses one date format across the entire slip.

None of the five is conclusive on its own. Combined with at least one of signs 1-6, they shift the file from review to rejection. The visual tells are evidence; the cross-system tells are proof.

Catalogue the limits of AI fabrication and the verification workflow becomes clearer. AI is now reliable at:

• Layout fidelity — column widths, row spacing, fonts, basic table structure.

• Internal arithmetic — basic plus allowances equals gross, gross minus deductions equals net.

• Plausible-sounding employer names, employee codes, addresses, and bank narrations.

• Plausible TDS amounts at the right percentage of gross.

• Watermarks, company logos pulled from public sources, signatory designations.

• HRA exemption arithmetic that conforms to Section 10(13A) and Rule 2A.

AI is poor at, or systematically incapable of:

• Producing a slip whose claimed TDS deposit exists in Form 26AS under the named employer's TAN.

• Producing a slip whose claimed salary credit exists in AIS under the same PAN.

• Producing a UAN that resolves to a real EPFO member record matching the named employee.

• Producing an employer name that exists on the MCA registry with Active status and an authorised-signatory name that matches the company's actual Director or KMP list.

• Producing twelve months of slips whose corresponding bank credits show natural rupee-level variance.

• Producing a valid verification QR. The QR is bound to the specific document content at the moment of legitimate generation; any later edit to the document invalidates the QR.

The verification workflow targets every line in the second bucket. Visual checks (signs 1-7 above) are the fast filter; the institutional cross-checks (26AS, AIS, EPFO, MCA) are the conclusive filter; the verification QR short-circuits both filters when present.

AI fabricates pixels. AI cannot fabricate a valid verification QR. The verification QR is bound to the document at the moment of legitimate generation: when an HR team issues a salary slip through the salary slip generator, the document carries a QR alongside a verifier URL. Any party scanning the QR can confirm in seconds whether the document is unmodified since issue. The check displaces font comparison, arithmetic spot-checks, and most of the visual-tell scan for that document.

The limitation is that verification QRs only cover documents whose issuer chose to add one. An applicant submitting a salary slip without a verification QR still requires the full seven-sign + cross-check workflow. Adoption matters: the more employers issue verification-strengthened slips, the larger the share of files that resolve via the QR scan and the more attention HR/lender teams can give to the residual unscanned files.

For enterprises issuing salary slips, payment receipts, GST invoices, and rent receipts to employees and vendors, the corporate bundle provides verification-strengthened generators across the document suite. For HR teams verifying incoming Form 12BB submissions, scanning the QR on any verification-strengthened rent receipt or payment receipt completes the authenticity check before the manual HRA-arithmetic step begins. For the upstream rent-side audit chain that pairs with the salary slip in any HRA file, see why payment receipts alongside rent receipts strengthen every HRA claim.

A six-step workflow scales to a high-volume HR or underwriting team and catches the bulk of AI fabrications in under three minutes per file:

1. Scan the salary slip for any verification QR. Scan it through the verifier. If the QR validates, the document is authentic; proceed to the HRA-arithmetic step. (15 seconds.)

2. Open the slip and run the visual-tells scan (signs 1-7 above). Net pay variance across months, UAN format, signatory designation, font and alignment, watermark layering. Note any flags. (45 seconds.)

3. Look up the employer name on the MCA master-data search. Confirm status is Active and the entity type matches the signatory designation on the slip. (30 seconds.)

4. If the slip carries a GSTIN, validate it on the GST taxpayer search. Confirm the GSTIN resolves to the same entity name as the slip. (30 seconds.)

5. Pull or request the applicant's most recent Form 26AS and confirm that quarterly TDS deposits exist under the employer's TAN at the rate the slip claims. (45 seconds, assuming the applicant has 26AS ready; longer if it requires a fresh pull.)

6. If steps 1-5 produced any unresolved flag, escalate to a manual review with a bank statement request. Else, proceed to apply the HRA exemption under Rule 26C and move to the next file. (15 seconds.)

The workflow front-loads the cheapest checks (QR scan, visual scan) and reserves the slowest (26AS pull, bank statement request) for files that need them. At a 1,000-file daily volume, the verification QR short-circuit takes 10-20 percent of files off the manual scan entirely; the remaining 80-90 percent move through the three-minute scan. Where verification QR adoption rises in a particular employer base, the team's effective throughput rises with it.

For HR teams issuing slips, payslips, and Form 16 themselves and looking to embed the QR in outgoing documents, the corporate bundle is the operative path; for self-service salaried employees and small-employer HR running ad-hoc, the salary-slip generator issues a single verification-strengthened slip in under 60 seconds without account creation.